Langfuse is an open-source platform for tracing, evaluating and managing prompts in LLM apps, which teams can use in the cloud or host themselves.
In plain words
Langfuse records traces of what an app does, including model calls, retrieval, embeddings and agent actions, so you can see where things go wrong. It also runs evaluations, manages and versions prompts, and supports experiments against datasets. Its code is open, and you can run it yourself, locally, in your cloud or on-premises. Since January 2026 the company has been part of ClickHouse.
A quick example
Imagine a team whose chatbot got slower after a change. In Langfuse they open recent traces, see that the retrieval step now takes most of the time, fix it, and run an experiment on a saved dataset to check answer quality did not drop.
Where you'll meet it
- Observability for LLM apps and agents.
- Teams that need to self-host their tracing data.
Related
Sources
- Docslangfuse/langfuse, Langfuse (GitHub) · read 29 Sept 2026
- DocsLangfuse documentation, Langfuse · read 29 Sept 2026
- DocsSelf-host Langfuse, Langfuse documentation · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 11 Nov 2026. Spotted an error? Report it.